On Fri, 12 Sep 2008, David Cantrell wrote:

On Thu, Sep 11, 2008 at 08:52:33PM +0100, Tom Anderson wrote:
On Thu, 11 Sep 2008, David Cantrell wrote:
The latter would mean making cyclists identifiable, with number plates.
Good idea.
Only if you wanted to do it with cameras. If there were actual people,
they could stop them there and then.

Good luck catching a cycle courier when you're on foot!

Taser!

Have you
come across this gait analysis business, for instance? Basically, software
can extract enough unique information about the way someone walks to
uniquely identify them in a crowd. Big-brother-tastic! Detecting dangerous
lane changes would seem trivial by comparison.

A lane change is only dangerous by nature of the other traffic around
it. To spot a dangerous lane change you need to "understand" the whole
picture.

Yes. And i don't think that's beyond the abilities of a computer. It won't
be as good as a human, but if it can be 80% as good for 1% of the cost,
then that's a win.

And if you're going to have a person, why not just have them stand next
to the junction in question?

Well, because one person can monitor more than one camera.

Not very well. And even if they can, once they've spotted a cyclist (or
a pedestrian, or a horse rider, or a driver) being naughty they thenr
have to dispatch someone to go and nick 'em, by which time it's too
late.

if it's a driver, they just note the plate and send them a fine.

Non-plated road users can't be caught in this way, but then they account
for a tiny fraction of dangerous road use.

tom

--
Subvert the paradigm!